Sign up ×
Physics Stack Exchange is a question and answer site for active researchers, academics and students of physics. It's 100% free.

Is it true that when you take the covariant derivative of a covariant tensor, do you always have to do with a subscript? What if you do it wrt a superscript?Does the first term (with the partial derivative) take a minus sign? More specifically, is this true?

$$\nabla^{\mu}R_{\mu\nu} = -\frac{\partial R_{\mu\nu}}{{\partial x_{\mu}}} + \text{(Christoffels)}$$

Where does the minus sign come from? Is there a proof for this, or is it just a definition? Also, is there a change in the signs for Christoffel symbols(not the change if the the tensor's indices change position, but the change when the index of the covariant differential changes)?

I want to know the PROOF/REASON behind the minus sign.

share|cite|improve this question
Remember that $x^\mu x_\mu = x_\mu x^\mu$. In other words, if the subscript of the covariant derivative is contracted, then it doesn't really matter if it is an upper index provided that the other (contracted) index is lowered (and vice versa). Also, the minus sign you are referring to only happens when you raise/lower the time component (or spatial component, depending on the metric signature you are using). – Hunter Jun 9 '14 at 15:30
The metric signature is (-+++), and I read this thing on the internet. He didnt explain how did he get it, so I got confused.. – GRrocks Jun 9 '14 at 16:09
"I read this thing on the internet"... What exactly did you read on the internet, i.e. what exactly confuses you? – Hunter Jun 9 '14 at 16:12
The reason you have a minus sign there.....And here is the link… – GRrocks Jun 9 '14 at 16:14

2 Answers 2

No. The subscript is the defined thing. If you have the superscript, you just assume raising with the metric tensor, so:

$$\nabla^{\mu}R_{\mu\nu} \equiv g^{\mu\alpha}\nabla_{\alpha}R_{\mu\nu}$$

which you expand normally with partial derivatives and Christoffels. Of course, since we know that $\nabla^{a}\left(R_{ab} - \frac{1}{2}Rg_{ab} \right)= 0$, we know right away that we can simplify $\nabla^{\mu}R_{\mu\nu}$ to $\frac{1}{2}\nabla_{\nu}R$

share|cite|improve this answer
thanx for the answer @jerry Schirmer...however, Can you explain where the minus sign comes from(I know the Bianchi identities, but was trying to derive the LHS of the field equations in the manner einstein might have, as he didnt have those identities....) – GRrocks Jun 9 '14 at 16:12
@GRrocks: why do you dismissively say that the Bianchi identities wheren't around? I can't say for sure that they weren't, but most of Luigi Bianchi's work predates 1916: – Jerry Schirmer Jun 4 at 18:57
@Jerruy Schirmer I really don't think Einstein would have used the Bianchi identities. I am not really into biographies, but I read one and it said that Einstein first gave the field equations only in terms of the Ricci tensor, and later realised the mistake and changed the equations to what we now study. He would'nt have done it in the first place if he knew the identities, would he? I was just toying around with the idea that how would he have done it...... – GRrocks Jun 6 at 10:53
@GRrocks: if he didn't use the general theorem, my guess would be that he brute force derived the identity $\nabla^{b}R_{ab} = \frac{1}{2}\nabla_{a}R$ – Jerry Schirmer Jun 6 at 14:56
hehe....maybe....wouldn't be surprised if he did it though....... – GRrocks Jun 9 at 9:10

From your comments, I will try to answer what confuses you. Let us take a metric signature: $$ \eta_{\mu \nu} = \mathrm{diag}(-1,1,1,1) $$ and let us consider some general $x^\mu$. We will denote the time component of $x^\mu$ by $x^0$. If we want to lower the index of $x^0$, we get: $$x^0 = \eta^{0 \mu} x_\mu = \eta^{0 0} x_0 + \eta^{0 1} x_1 + \eta^{0 2} x_2 + \eta^{0 3} x_3 \tag{1}$$ Since $\eta^{0 0} = -1$ and $\eta^{0 1} = \eta^{0 2} = \eta^{0 3} = 0$, equation $(1)$ becomes: $$x^0 = - x_0$$ and so we get the minus sign.

Note that if we only consider the spatial component $x^i$ (where $i$ is either the $1$st, the $2$nd or the $3$rd component), then we lower the index again as: $$ x^i = \eta^{i \mu} x_\mu = \eta^{i0} x_0 + \eta^{i1} x_1 + \eta^{i2} x_2 + \eta^{i3} x_3 = x_i$$ and so we don't get a minus sign.

share|cite|improve this answer
OKKK, so the minus sign is really because we are lowering the indices of the covariant derivative wrt the MINKOWSKI metric..But would this be valid if we did it wrt the METRIC TENSOR for ANY line element?? – GRrocks Jun 9 '14 at 16:24
Yeah, for any Lorentzian metric in its canonical form this will be valid (remember it always possible to put a metric into its canonical form for some point on the manifold). What kind of metric do you have in mind? – Hunter Jun 9 '14 at 16:41
Thanx @Hunter, I was just thinking of something arbitrary(e.g swarzchild) – GRrocks Jun 9 '14 at 17:01
I'm not sure, but you'll probably have to look into curvilinear coordinates; remember that the basis of the tangent and its cotangent space must satisfy $e^\mu e_\nu = g_{\mu \nu} e^\mu e^\nu = \delta^\mu_\nu$. – Hunter Jun 9 '14 at 17:35
@GRrocks: you can always diagonalize the metric at a point by just defining new coordinates $z^{a} = \lambda^{a}{}_{b}x^{b}$ for some set of constant $\lambda$. – Jerry Schirmer Jun 9 '14 at 18:29

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.